Postman vs Old Tools: What Changed for Data Scientists.

Old Way: Hitting a data API with curl

# Terminal
curl -X POST https://api.example.com/v1/extract \
  -H "Authorization: Bearer $TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"query": "SELECT * FROM users", "limit": 1000}' \
  -o output.json

# Then load in Python
import requests
import pandas as pd

headers = {"Authorization": f"Bearer {TOKEN}"}
r = requests.post(
    "https://api.example.com/v1/extract",
    headers=headers,
    json={"query": "...", "limit": 1000}
)
df = pd.DataFrame(r.json()["rows"])

Pain points:
  - Token management is manual
  - No history of what you tried
  - Every test = kernel restart
  - No team sharing

Postman Way: Same API in a structured collection

1. Create Collection: "Data Extraction APIs"
2. Add Request: POST /v1/extract
   - Auth: Inherited from collection (Bearer)
   - Body: {"query": "...", "limit": 1000}

3. Add Pre-request Script (runs automatically):
   pm.environment.set("token", getNewToken());

4. Add Tests (runs after response):
   pm.test("Status is 200", () => {
     pm.response.to.have.status(200);
   });
   pm.test("Has rows", () => {
     pm.expect(pm.response.json().rows.length).to.be.above(0);
   });

5. Use Environments to switch dev/prod

6. Share with team via workspace

7. Run from CI with Newman:
   newman run data-collection.json \
     -e production.postman_environment.json

Benefits:
  - Token auto-managed
  - History saved
  - Repeatable and shareable
  - CI-ready

4-Week Postman Adoption Plan:

Week 1: Basics
  - Install Postman desktop app
  - Send your first GET/POST request
  - Save requests to a collection
  - Explore response views (Pretty, Raw, Preview)
  - Goal: First collection with 5-10 API calls

Week 2: Environments + Auth
  - Create environment variables ({{base_url}}, {{token}})
  - Set up API Key / Bearer / OAuth2
  - Switch between dev/prod environments
  - Goal: Auth working across environments

Week 3: Tests + Scripts
  - Add tests to verify status codes
  - Test JSON response structure
  - Write pre-request scripts for tokens
  - Chain requests (extract ID from response A, use in request B)
  - Goal: Collection with automated tests

Week 4: Newman + Collaboration
  - Install Newman (npm install -g newman)
  - Run collection from command line
  - Integrate with GitHub Actions
  - Share collection with team via workspace
  - Goal: CI-integrated API test suite

Total: ~6-8 hours of learning over 4 weeks
Result: Confident, structured API workflow for ML/DS projects

Data Science Use Cases for Postman:

1. Data Collection from APIs
   - REST APIs for weather, finance, social, etc.
   - Use environments for API keys
   - Save responses for later analysis

2. Model Inference Testing
   - Test model serving endpoints (FastAPI, Flask)
   - Verify predictions match expected outputs
   - Set up automated regression tests

3. Feature Store APIs
   - Query feature stores (Feast, Tecton)
   - Test batch and online features

4. MLOps Pipelines
   - Trigger training jobs via REST
   - Check pipeline status
   - Get model performance metrics

5. Third-Party Integration
   - OpenAI, Hugging Face, Cohere APIs
   - Test prompt responses before coding
   - Save example requests for team

6. Debugging Production Issues
   - Reproduce issues reported by stakeholders
   - Check response headers, timing, errors
   - Share bug reproduction with backend team

Each use case becomes a reusable collection.